> The $20-200 LLM plans are all subsidized and aren't paying for themselves. Something has to give here.
Whats interesting to me as well as much as companies are pushing AI adoption, i have started to hear AI token spend limits enforced across a few companies, so its not entirely clear that b2b can make them profitable yet either.
If all the models reach good enough, then low cost provider would win. Gemini seems like a safer bet since Google controls more of the stack / has more efficiencies / cross selling / etc.
It’s not like “best” has won any other b2b arms race in the past.
>If all the models reach good enough, then low cost provider would win. Gemini seems like a safer bet since Google controls more of the stack / has more efficiencies / cross selling / etc.
Gemini is the best deal too. For $20: you get multiple quotas per day across the products (web, CLI, antigravity, AI Studio) 2tb of cloud storage, and you can family share the plan.
I don't know Gemini's pricing model in detail, but in general pricing doesn't generalize well between personal/hobbyist and enterprise use. Consumer pricing of variable costs is a balancing act, and most Gemini users aren't going to be anywhere near the quota; a company of 1000 can't always buy for $20,000 what 1000 random users with $20 personal plans are theoretically capped at.
Ultimately though in the long run..
They invented the tech, have a large cashflow generating business subsidizing R&D as well as sales, with network effect of existing B2B relationships.
Further they have their own TPUs, datacenters, etc on which to run their models.
Plus existing data they've squirreled away over the preceding 30 years from books, web, etc.
Just seems like a lot of efficiencies if its going to come down to cost.
In large part because most companies have a set budget for IT spend. Thats how “normal” profitable companies operate outside this cash burning bonanza that’s going on.
And in that reality one can’t just magically spend a bunch more on some fancy new thing, especially when said fancy new thing isn’t retuning value. So “token limits” and cost controls on B2B is entirely expected here.
> especially when said fancy new thing isn’t retuning value
I think this is the key element. Either they can't measure the value, or it's far far lower than anyone wants to believe, or both.
I think the problem is less that it makes some coding tasks XX% faster, but that the end to end of a SWEs roles tasks is only improved by some much smaller Y%.
If a CTO sets $10k/year spend limits on $500k SWEs.. they must not believe any of the hype.
The problem is that AGI fantasy aside, CTOs at companies are expected to deliver results today and tomorrow. Better to let somebody else hold the bag and train models, then once it finally works as advertised you can ease on the brakes.
> Just take a look at the enterprise. Amazon's ad business alone is already a better business than Oracle or SAP or Salesforce, with superior margins, and it's growing faster too.
You can say the same about AWS and then prove the b2b case instead of ad case as well
AWS is legitimately a giant and it should be considered in enterprise broadly. It's infrastructure more than enterprise software of course, which is where Anthropic is at. Anthropic is not trying to host the world's databases and services (at present anyway). Anthropic will however help you write software to compete with Salesforce, Oracle, SAP, et al.
Google's ad business remains far larger and more profitable than AWS. And the advertising segment is drastically larger than the segment AWS is in. Just Google + Meta = nearing $600 billion in ad sales. Amazon will soon have their own $100 billion in ad sales.
I guess the question is how many more $100B of ad sales slots are available, aside from just stealing share from incumbents (who already took it from traditional media channels over last 20 years).
At some point someone needs to add value to the real economy, not just take an ad tax off the top.
I don't think advertising is entirely a zero-sum game (there is real value to everyone for steveBK123 to learn of a product that actually solves a real need or saves money, etc) - but it has to be something akin to it - the economy can't support five hundred trillion dollars in advertising spend.
Yes, i think the Trump admin has escalated itself into a situation that either involves ground troops or leaving without opening the strait.
The first is bad due to the losses that will be incurred and the difficulty of holding territory.. for unclear strategic reasons (I thought we destroyed their nuclear program last summer / what was the urgency / is this even our war?). The second is bad because the strait was open before this started, so things are worse than starting conditions.
That is not to say Iran is winning. Remember this is not a sports game, and no one needs to win. It is possible, and likely, for everyone to lose (be in a worse position than prior).
> either involves ground troops or leaving without opening the strait.
These options are not mutually exclusive.
> That is not to say Iran is winning.
They are though, the US administration has already lost it's patience, their strategic objectives (whatever they might have been have clearly not materialized), the talk about talks may very well be the administration preparing to make a bunch of concessions proclaim victory and walk away.
As it's possible for both parties to lose, a party can win all the battles and lose the war.
It is hard to game out the best scenario here. Wait, it really isn't. We should just stop. Make a deal with Iran, accept egg on our face and step back. Why? Because they are destabilized. They are likely to crumble. If we keep attacking then they stay alive. If we go away then they have to deal with their broken infra and deeply unhappy population. They were on the path until we hit them. Then, like nearly every country ever, it gave their government legitimacy. If we walk away and focus, hard, on helping the gulf nations that we just hurt badly it will stabilize the region and allow them to fall. But that will never happen because we went into this due to ego and we will stay due to ego.
What if Iran escalates when US decides to go? I don’t think US can go without leaving a power vacuum, which, given current forces positioning, would benefit Iran most probably. I don’t see a path to helping Gulf nations, which will pragmatically be inclined to work with Iran as neither of them can leave like US can.
>That is not to say Iran is winning. Remember this is not a sports game, and no one needs to win. It is possible, and likely, for everyone to lose (be in a worse position than prior).
As of right now, Iran looks likely to end the war with permanent control of the strait of Hormutz. They'll tax the gulf countries in perpetuity.
Gulf countries can't reasonably afford to go to war with Iran over this either, and it's even less likely that they could prevail in such a conflict. Gulf countries can't even afford to go to war with Iran now, with the US actively fighting there.
Iran can suffer terrible short-term and medium-term economic consequences while still establishing a whole new kind of dominance over the region.
The problem is that we need to adapt to the asymmetrical aspect of drone warfare, as Ukraine has done. The best description I saw of the current state is “flying IEDs”.
Drones and ballistic missiles make area denial asymmetrically cheap for a defending forces. This lesson needs to be incorporated because it would be the same tactic used by China to deny access to the South China Sea.
> The longer your product exists the more important the quality of the code will be. This obsession so many have with "get it out the door in 5 seconds" is only going to continue the parade of garbage software that is slow as a dog, and uses gigabytes of memory to perform simple tasks.
Exactly. A lot of devs optimizing for whether the feature is going to take a day or an hour, but not contemplating that it's going to be out in the wild for 10 years either way. Maybe do it well once.
> but not contemplating that it's going to be out in the wild for 10 years either way
I think there are a lot of developers working in repos where it's almost guaranteed that their code will _not_ still be there in 10 years, or 5 years, or even 1 year.
>I think there are a lot of developers working in repos where it's almost guaranteed that their code will _not_ still be there in 10 years, or 5 years, or even 1 year.
And in almost all of those cases, they'd be wrong.
I think I calculated the half-life of my code written at my first stint of Google (15 years ago) as 1 year. Within 1 year, half of the code I'd written was deprecated, deleted, or replaced, and it continued to decay exponentially like that throughout my 6-year tenure there.
Interestingly, I still have some code in the codebase, which I guess makes sense because I submitted about 680K LOC (note: not all hand-authored, there was a lot of output from automated tools in that) and 2^15 is 32768, so I'd expect to have about 20 lines left, which is actually surprisingly close to accurate (I didn't precisely count, but a quick glance at what I recognized suggested about 200 non-deprecated lines remain in prod). It is not at all the code that I thought would still be there 15 years later, or that I was most proud of. The most durable change appears to be renaming some attributes in a custom templating language that is now deeply embedded in the Search stack, as well as some C++ code that handles how various search options are selected and persisted between queries.
I think this both proves and disproves the original point. Most of your code is temporary. You have no idea which parts of your code is temporary. It's probably not the parts that you wish were temporary, which will almost certainly be made permanent.
In my experience the code will, but by year 5 nobody is left who worked on it from inception, and by year 10 nobody knows anybody who did, and during that time it reaches a stage where nobody will ever feel any sense of ownership or care about the code in its entirety again.
I come into work and work on a 20 year old codebase every day, working on slowly modernizing it while preserving the good parts. In my experience, and I've been experimenting with both a lot, LLM-based tools are far worse at this than they are at starting new greenfield projects.
When it comes to professional development, I've almost never worked on a codebase less than 10 years old, and it was always [either silently or overtly] understood that the software we are writing is a project that's going to effectively live forever. Or at least until the company is no longer recognizable from what it is today. It just seems wild and unbelievable to me, to go to work at a company and know that your code is going to be compiled, sent off to customers, and then nobody is ever going to touch it again. Where the product is so throwaway that you're going to work on it for about a year and then start another greenfield codebase. Yet there are companies that operate that way!
How can you possibly know which type of repo you're in ahead of time? My experience is that "temporary" code frequently becomes permanent and I've also been on the other side of those decisions 40 years later.
Unless you’re producing demos for sales presentation (internally or externally), it’s always worth it to produce something good. Bad code will quickly slow you down and it will be a never ending parade of bug tickets.
That depends on how quick the feedback loop is for your decisions. If it takes weeks or months to find the impact of your changes, or worse, if you're insulated somehow from those changes, you may not be pushed toward improving the quality of your code.
I think type1 vs type2 dev requirements are also dependent on lifecycle / scale of your project, not just that its library / framework / mission critical software.
If you aren't even sure if your idea is even gonna work, whether you have PMF, or the company will be around next year.. then yeah.. speed over quality all day long.
On the other hand, I've never done the startup thing myself, and tend to work on software project with 10-20 year lifecycles. When code velocity maximalism leads to outages, excess compute cost and reputational issues.. good code matters again.
Re: "No one has ever made a purchasing decision based on how good your code is."
Sonos very much could go out of business for agreeing with this line. I can tell you lots of people stopped buying their products because of how bad their code quality became with the big app change debacle. Lost over a decade of built up good will.
Apple is going through this lately with the last couple major OS releases across platforms and whatever is going on with their AI. This despite having incredible hardware.
I've mentioned this in another thread when the topic came up, but this has other effects like availability of backup childcare.
With kids, unbelievably, needing some form of car seat until the age of 8~12.. you are limited in who can help with pickup/dropoff/after school programs to someone who borrows your car or is doing so often enough to buy the correct sized car seats for each of your kids. The upper limit being 12 is pretty wild considering we then let them drive at 16.
Oddly of course in the US, if your kid is taking the school bus they are entirely unbelted and without a safety seat from the age of.. 5.
As an uncle to multiple kids under 10, I end up only spending time with them in their own home or when they are brought over.
Contrast that with my upbringing spending time with cousins/aunts/neighbors being driven around by whoever I was staying with for the afternoon.
In the UK, children that would otherwise need a child seat can travel without one in a taxi, or if the journey is unexpected and necessary, or if your car does not have rear seat belts.
So I guess you need to get something from the early 80s.
If they're too short to fit safely in a seat and use the normal seat belt without one, then yes, a 12-year-old needs a booster seat even if it's not a "child seat".
In the UK the limit is 135cm, which is about the right height for the average 9-year-old, so you're talking about a very small 12-year-old indeed - a good 15cm shorter than average.
I've observed that a whole lot of people absolutely do not keep putting kids in car seats once they're about the height & weight of a petite adult (us included).
So the guidelines say one thing, but I'd be surprised if a majority of parents are still putting their kids in them even by age 10, let alone 12.
Yes it’s final form of the evolution that social media started.
Village idiot used to be found out because no one in the village shared the same wingnut views.
Partisan media gave you two polls of wingnut views to choose for reinforcement.
Social media allowed all village idiots to find each other and reinforce each others shared wingnut views of which there are 1000s to choose from.
Now with LLMs you can have personalized reinforcement of any newly invented wingnut view on the fly. So can get into very specific self radicalization loops especially for the mentally ill.
But also Musk needs to get paid $1T, and he also needs indices to change their rules to pump more of your money into his giga-IPO.
Nothing to see here.
reply